【kafka KSQL】遊戲日誌統計分析（2）

接上一篇文章【kafka KSQL】遊戲日誌統計分析（1），展現一下KSQL WINDOW 功能的使用。

測試用日誌數據：

{"cost":7, "epoch":1512342568296,"gameId":"2017-12-04_07:09:28_高手2區_500_015_185175","gameType":"situan","gamers": [{"balance":4405682,"delta":-60,"username":"lza"}, {"balance":69532,"delta":-60,"username":"lzb"}, {"balance":972120,"delta":-60,"username":"lzc"}, {"balance":23129,"delta":180,"username":"lze"}],"reason":"game"}

KSQL三種Window

統計每2分鐘內完成對局大於等於3局的玩家

• 根據時間窗口（Tumbling window）創建table：

CREATE TABLE users_per_minute AS \
    SELECT username, COUNT(*) AS game_count, SUM(delta) AS delta_sum, SUM(tax) AS tax_sum , WINDOWSTART() AS win_start, WINDOWEND() AS win_end \
    FROM USER_SCORE_EVENT \
    WINDOW TUMBLING (SIZE 2 MINUTE) \
    WHERE reason = 'game' \
    GROUP BY username;

• 過濾出game_count大於3局的玩家：

SELECT username, game_count, win_start, win_end FROM users_per_minute WHERE game_count >= 3;

輸出：

lze | 6 | 1546744320000 | 1546744440000
lzc | 6 | 1546744320000 | 1546744440000
lza | 6 | 1546744320000 | 1546744440000
lzb | 6 | 1546744320000 | 1546744440000
lzb | 3 | 1546744440000 | 1546744560000
lzc | 3 | 1546744440000 | 1546744560000
lza | 3 | 1546744440000 | 1546744560000
lze | 3 | 1546744440000 | 1546744560000

統計曾在10分鐘以內完成過3局牌局的玩家

不限定某個特定的10分鐘，只要在某個10分鐘以內完成了便可。

• 建立HOPPING WINDOW時間窗口Table：

CREATE TABLE users_hopping_10_minute AS \
    SELECT username, COUNT(*) AS game_count, SUM(delta) AS delta_sum, SUM(tax) AS tax_sum , TIMESTAMPTOSTRING(WindowStart(), 'yyyy-MM-dd HH:mm:ss') AS win_start, TIMESTAMPTOSTRING(WindowEnd(), 'yyyy-MM-dd HH:mm:ss') AS win_end \
    FROM USER_SCORE_EVENT \
    WINDOW HOPPING (SIZE 10 MINUTE, ADVANCE BY 30 SECONDS) \
    WHERE reason = 'game' \
    GROUP BY username;

• 過濾出game_count大於等於3的玩家

SELECT username \
FROM users_hopping_10_minute \
WHERE game_count >= 3 \
GROUP BY username;